Recently the SDR# team have updated the algorithm on the noise reduction plugins used in SDR#. It appears that both the IF and Audio noise reduction plugins were updated with a better smoothing algorithm. We briefly tested the new algorithm and compared it against an older version. The new algorithm has noticeably less hiss and is slightly clearer when compared at the same noise reduction level. We tested with the same threshold levels and using the speech profile.
At the same time we've also seen news that Simon of SDR-Console is working on another noise reduction algorithm based on deep neural networks in the latest private beta version. A video of it in action was posted by Paul J in the SDRplay users group (note that you will need a Facebook account and will probably need to be a member of the SDRplay group to view that video). The algorithm seems to be based on the RNNoise paper that was posted here. The SDR# algorithm was also tweaked based on information gained from that paper although it doesn't use neural networks directly.
SDR# NR Comparison
Old SDR# NR Algorithm
New SDR# NR Algorithm